split plot in time and space and combined analyses€¦ · split plot in time and the nested block...
TRANSCRIPT
SPLIT PLOT IN TIME AND THE NESTED BLOCK ARRANGEMENTS
Split Plot in Time
This arrangement occurs when you have an experiment where you collect data from
the same experimental unit over a series of dates.
An example of this would be an experiment that includes a perennial species (e.g.
alfalfa) and different harvest dates or an experiment where you measure water use
each week in a field experiment.
The ANOVA for this situation is similar to a split block.
Example
An experiment comparing the yield of 10 alfalfa cultivars cut at three different times
during the growing season. Cultivars and cutting times are fixed effects.
SOV Df Expected Mean Squares F-test
Rep r-1
2222
Rcttc
Cultivar
c-1
1
2
22
c
rtt
C MS/Error(a) MS
Error(a) (r-1)(c-
1)
22
t
Time
t-1
1
2
22
trcc
T MS/Error(b) MS
Error(b) (r-1)(t-1)
22
c
Cultivar*Time
(c-1)(t-1)
)1)(1(
)( 2
2
tc
r
C*T MS/Error(c) MS
Error(c) (r-1)(c-
1)(t-1)
2
Total rct-1
LSD’s for Split Plot in Time Arrangement
1. To compare two A means averaged over all time treatments (.e.g. a0 vs a1)
ta /2, Error(a) df
2Error(a)MS
r * time
= 4.3032(2.5150)
3X4
= 2.79
2. To compare two time means averaged over all A treatments (.e.g. time0 vs time1)
59.1
23
)412.0(2303.4
)(2t df Error(b) /2,
X
ra
MSbError
3. To compare two levels of A means at the same level of the time factor (e.g.
a0time0 vs a1time0)
072.3
43
]515.2794.0)14[(23.4LSD and
4.3515.2794.0)14(
)303.4(515.2)447.2)(794.0)(14(
)()()1(
)()()()()1(
*
])()()1[(2t
'
)(,2/)(,2/
'
'
x
t
MSaErrorMScErrortime
tMSaErrortMScErrortimet
and
timer
MSaErrorMScErrortime
ac
dfaErrordfcError
ac
ac
4. To compare two time means at the same level of factor A (e.g. a0time0 vs a0time1)
551.1
23
]412.0794.0)12[(2447.2LSD and
447.2412.0447.2)12(
)447.2(412.0)447.2)(794.0)(12(
)()()1(
)()()()()1(
])()()1[(2t
'
)(,2/)(,2/
'
'
x
t
MSbErrorMScErrora
tMSbErrortMScErrorat
and
ra
MSbErrorMScErrora
bc
dfbErrordfcError
bc
bc
5. To compare two vertical means at different levels of the horizontal factor (e.g. a0b0 vs
a1b3)
294.2
423
]515.2)12(412.0)14(794.0)14)(12[(2208.3LSD and
208.31515.2)12(412.0)14(447.2)14)(12(
515.2)12()447.2(412.0)14()447.2)(794.0)(14)(12()
)()1()()1()()1)(1(
)()()1()()()1()()()1)(1(
**
])()1()()1()()1)(1[(2t
'
)(,2/)(,2/)(,2/
'
'
xx
t
MSaErroraMSbErrortimeMScErrortimea
tMSaErroratMSbErrortimetMScErrortimeat
and
timear
MSaErroraMSbErrortimeMScErrortimea
abc
dfaErrordfbErrordfcError
abc
abc
SAS Commands for an RCBD with a Split Plot in Time Arrangement
options pageno=1;
data spplttim;
input A $ time $ Rep Trt Yield;
datalines;
a0 b0 1 1 13.8
a0 b1 1 2 15.5
a0 b2 1 3 21
a0 b3 1 4 18.9
a1 b0 1 5 19.3
a1 b1 1 6 22.2
a1 b2 1 7 25.3
a1 b3 1 8 25.9
a0 b0 2 1 13.5
a0 b1 2 2 15
a0 b2 2 3 22.7
a0 b3 2 4 18.3
a1 b0 2 5 18
a1 b1 2 6 24.2
a1 b2 2 7 24.8
a1 b3 2 8 26.7
a0 b0 3 1 13.2
a0 b1 3 2 15.2
a0 b2 3 3 22.3
a0 b3 3 4 19.6
a1 b0 3 5 20.5
a1 b1 3 6 25.4
a1 b2 3 7 28.4
a1 b3 3 8 27.6
;;
ods rtf file='example.rtf';
run;
proc anova;
class rep a time;
model yield=rep a rep*a time rep*time a*time;
test h=a e=rep*a;
test h=time e=rep*time;
means a/lsd e=rep*a;
means time/lsd e=rep*time;
means a*time;
title 'ANOVA for an RCBD with a Split Plot in Time
Arrangement';
run;
ods rtf close;
run;
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
Class Level Information
Class Levels Values
Rep 3 1 2 3
A 2 a0 a1
time 4 b0 b1 b2 b3
Number of Observations Read 24
Number of Observations Used 24
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
Dependent Variable: Yield
02:15 Sunday, June 17, 2012 205
Source DF
Sum of
Squares Mean Square F Value Pr > F
Model 17 511.3587500 30.0799265 37.91 0.0001
Error 6 4.7608333 0.7934722
Corrected Total 23 516.1195833
R-Square Coeff Var Root MSE Yield Mean
0.990776 4.298913 0.890771 20.72083
Source DF Anova SS Mean Square F Value Pr > F
Rep 2 7.8658333 3.9329167 4.96 0.0536
A 1 262.0204167 262.0204167 330.22 <.0001
Rep*A 2 5.0358333 2.5179167 3.17 0.1148
time 3 215.2612500 71.7537500 90.43 <.0001
Rep*time 6 2.4775000 0.4129167 0.52 0.7767
A*time 3 18.6979167 6.2326389 7.85 0.0168
Tests of Hypotheses Using the Anova MS for Rep*A as an Error Term
Source DF Anova SS Mean Square F Value Pr > F
A 1 262.0204167 262.0204167 104.06 0.0095
Tests of Hypotheses Using the Anova MS for Rep*time as an Error Term
Source DF Anova SS Mean Square F Value Pr > F
time 3 215.2612500 71.7537500 173.77 <.0001
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
02:15 Sunday, June 17, 2012 206
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
t Tests (LSD) for Yield
02:15 Sunday, June 17, 2012 207
Note
:
This test controls the Type I comparisonwise error rate, not the
experimentwise error rate.
Alpha 0.05
Error Degrees of Freedom 2
Error Mean Square 2.517917
Critical Value of t 4.30265
Least Significant Difference 2.7873
Means with the same letter are
not significantly different.
t Grouping Mean N A
A 24.0250 12 a1
B 17.4167 12 a0
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
02:15 Sunday, June 17, 2012 208
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
t Tests (LSD) for Yield
02:15 Sunday, June 17, 2012 209
Note
:
This test controls the Type I comparisonwise error rate, not the
experimentwise error rate.
Alpha 0.05
Error Degrees of Freedom 6
Error Mean Square 0.412917
Critical Value of t 2.44691
Least Significant Difference 0.9078
Means with the same letter are
not significantly different.
t Grouping Mean N time
A 24.0833 6 b2
B 22.8333 6 b3
C 19.5833 6 b1
D 16.3833 6 b0
ANOVA for an RCBD with a Split Split Plot in Time Arrangement
The ANOVA Procedure
t Tests (LSD) for Yield
02:15 Sunday, June 17, 2012 204
Level of
A
Level of
time N
Yield
Mean Std Dev
a0 b0 3 13.5000000 0.30000000
a0 b1 3 15.2333333 0.25166115
a0 b2 3 22.0000000 0.88881944
a0 b3 3 18.9333333 0.65064071
a1 b0 3 19.2666667 1.25033329
a1 b1 3 23.9333333 1.61658075
a1 b2 3 26.1666667 1.95021366
a1 b3 3 26.7333333 0.85049005
Randomized Nested Block Arrangement
In some multifactor arrangements, the levels of one factor (e.g., factor B) are similar
but not identical for different levels of another factor (e.g., factor A).
An example would be the agronomic evaluation of genotypes with and without
resistance to a non-selective herbicide such as glyphosate or glufosinate in a single
experiment.
If you use a split plot arrangement with herbicide as the whole plot and genotype as
the subplot, all genotypes would get treated with the herbicide regardless if they are
resistant or non-resistant to the herbicide.
Example
Whole = herbicide rate (0 and X-rate)
Subplot = genotype (Resistant 1, Resistant 2, Susceptible 1 and Susceptible 2
X-rate 0-rate
Resistant 1 Susceptible 2
Resistant 2 Resistant 2
Susceptible 1 Susceptible 1
Susceptible 2 Resistant 1
ANOVA
Sources of variation degrees of freedom
Replicate r-1
Herbicide rate h-1
Error(a) (r-1)(h-1)
Genotype (g-1)
Herbicide rate x Genotype (h-1)(g-1)
Error(b) by subtraction
Total (r x h x g)-1
What are the consequences of using this arrangement?
An alternative to the split plot arrangement would be to block such that only the
resistant genotypes get sprayed with herbicide and the susceptible genotypes do not
get sprayed with herbicide.
X-rate 0-rate
Resistant 1 Susceptible 2
Resistant 2 Susceptible 1
This arrangement allows for valid comparisons between:
1. Genotypes nested within an herbicide level (e.g. Resistant 1 vs. Resistant 2
or Susceptible 1 vs. Susceptible 2).
2. Herbicide rate averaged across genotypes (e.g. X-rate vs. 0-rate).
Valid comparisons cannot be made between genotypes nested in different
herbicide rates (e.g. X-rate and Resistant 1 vs. 0-rate and Susceptible 2).
ANOVA
Sources of variation Degrees of
freedom†
Method for calculating the F-statistic
Replicate r-1 Replicate MS/Error MS
Herbicide h-1 Herbicide MS/Error MS
Genotype(Herbicide) (b-1)+(nb-1) Genotype(Herbicide MS/Error MS
Error (r-1)(G-1)
Total (r x G)-1
†r = number of replicates, h = number of herbicide treatments, b = number of biotech
(i.e., resistant) genotypes, nb = number of non-biotech (i.e., susceptible) genotypes, and
G = total number of genotypes in the experiment (i.e., b + nb).
Mean Separation
Valid mean separation can be done between:
1. Genotypes within a herbicide level (e.g. Resistant 1 vs. Resistant 2 or
Susceptible 1 vs. Susceptible 2).
The formula for calculating the least significant difference (LSD)
between genotypes within a class at p=0.05 is:
r
MSError x 2dfError ,2/05.0t
2. Herbicide rates averaged across genotypes (e.g., X-rate vs. 0-rate).
The formula for calculating the least significant difference (LSD)
between herbicide rates averaged across genotypes at p=0.05 is:
nbr x
1
br x
1MSError dfError ,2/05.0t
Where b = number of biotech (i.e., resistant) genotypes and
nb = number of non-biotech (i.e., susceptible) genotypes.